h a l f b a k e r ySuperficial Intelligence
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
One limitation of "green screen" photography is that the color of any partially-transparent pixel will be influenced by the color of the green screen. Digital matte shots could be made cleaner if they were done using a backdrop illuminated by a sodium lamps, and were shot using cameras with four CCD's--one
each for red, green, and blue (with filters to block out sodium light), and one for transparency (which would pick up nothing but sodium light). This would avoid color distortion of partially-transparent areas, and even allow photographing full-color translucent objects. Adding an extra imaging sensor would increase the cost of a camera, but I suspect that in many cases the optics on high-end cameras are a bigger factor in the cost than the sensors.
A slight variant on the 3D projection method I described.
http://www.faqs.org...nts/app/20090257120
[spidermother, Feb 17 2011]
Sodium Vapour Process
http://en.wikipedia...odium_vapor_process
It turns out this was baked, back in the pre-digital era. [Wrongfellow, Jan 19 2014]
[link]
|
| |
In theory a good idea - Sodium light is a single frequency so if you can filter out that frequency you'll be OK. I suspect though (*) that any filter is likely to filter out a broader range of frequencies and that this will impact the image quality, especially if you're talking about frequencies which occur in human skin tones.
(*) based on very little knowledge of the physics. |
|
| |
Actually, sodium light is mostly 2 slightly different frequencies. |
|
| |
Also, to get a really sharp line source, you have to have the vapour at a very low pressure, which affects the amount of light you get. |
|
| |
To get a useful quantity of light out of the lamp, you'll have to have a decent amount of vapour in there, which will broaden the emission line as the sodium atoms interact with each other. |
|
| |
You'll still get quite a well defined spectral peak out of a sodium lamp, though, and if you construct your 'white' light out of three similar peaks matched to the red, green and blue wavelengths that the camera's CCD responds to, you might be able to make this work. |
|
| |
I had an idea for a magic trick once, based on two "white" lights made up of different sets of spectral lines, and an object whose reflectivity at different wavelengths was such that it would appear one colour when lit by one "white" light and another colour when lit by the other - despite both lights appearing equally "white" to the human eye. |
|
| |
(Artists will sometimes tell you that there's no such thing as white - and they're correct.) |
|
| |
There's a method of 3D projection that uses that principle, [Wrongfellow]. The images for the left and right eye use different frequencies of red, green, and blue, with the intensities adjusted according to the eye's sensitivity curves, so that colours are rendered approximately correctly. The observer wears glasses that selectively admit those frequencies to the corresponding eye. The result is colour 3D, without the need for polarisation. |
|
| |
Which suggests that [supercat]'s idea could work. |
|
| |
LEDs are now available in a very large number of dominant frequencies (about 20 just from the type of semiconductor, hundreds if you include the manufacturers' frequency bins), so your magic trick would be quite bakable. |
|
| |
You could even use electronic controllers to trace out constant-apparent-white-point paths, so that the light would appear to remain white, while some objects illuminated by it would slowly change colour. |
|
| |
Oooh! Shiny! I'd definitely bun that if it was posted here. |
|
| |
I bet the design of the projector is interesting. |
|
| |
It's not for here, because it's been done. It uses 2 projectors, from memory, and software to do the necessary watchamacallit, colour space transforms. |
|
| |
(I would argue that an object that reflects close to perfectly at all visible frequencies is a white object, but would agree that there is no such thing as white in isolation). |
|
| |
LEDs are available in a wide range, but don't produce
single/narrow frequency output. Most have at least
a 30nm bandwidth, and some can be significantly
higher. It's fairly easy (although not especially
cheap) to get optical filters with blocking that will
only pass ~10-15nm bands though. You're still going
to get some crosstalk off your objects/reflective
surface from both fluorescence and Raman
scattering, although probably not enough to be visible . |
|
| |
//Most have at least a 30nm bandwidth// True, but that's about 1/3 - 1/4 of the sensitivity bandwidth of each type of cone cell. |
|
| |
(Frikkin') lasers would be ideal^H^H^H^H^Hbetter, but pricey. |
|
| |
//Sodium light is a single frequency// |
|
| |
I wonder if you'd get Raman scattering introducing side-
bands? With lasers it's a big problem. |
|
| |
How about using time resolution for the matting? Along the
lines of several threads recently, the two images could be
illuminated at high frequency but out of synch. |
|
| |
I think [bigsleep] gets what I'm aiming at. The idea isn't to produce something that looks RGB, or even any particular color. The purpose is to have a means of identifying which pixels can "see" through to the backdrop, and to what extent, without distorting the color of pixels that can slightly see through to the backdrop, thus allowing much cleaner mattes than would otherwise be possible. |
|
| |
[hippo]: People who work with hot glass (myself included) use glasses which didinium lenses which do a pretty good job of filtering out sodium light without affecting most other colors too badly. Color rendition of didinium lenses isn't perfect--one pale blue color of glass is clearly different from a pale violet when viewed with the naked eye in natural light, but indistinguishable when viewed through didinium. On the other hand, one could select colors for people's costumes, make-up, and sets which don't have that problem. |
|
| |
I'm not sure how well sodium-screen could be used to force an object which is in front of the backdrop to be transparent. I would guess that one might be able to do a reasonable job if one located a sodium lamp on the camera and covered objects to be obscured with front-projection film. All normal objects on the set would reflect the sodium light to some degree, but the objects which were covered with front-projection film would reflect it back toward the camera to a degree far greater than any object not so covered (the background screen would likely have to be covered with FP film as well). If the gain on the sodium sensor were low enough, the incidental light reflected off normal objects would be too dim to have any effect. |
|
| |